In the conventional art, a car navigation system, an automatic driving system, and the like are widely known as a system using a host vehicle position on a map. A global positioning system (GPS) is known as technology which is used for the systems.
In the convention system, since a host vehicle position on a map has an error of about several meters, for example, there is a problem in that an intersection in which the vehicle has to make a turn is mistaken or a vehicle travels on a lane which does not guide the vehicle to a destination.
In order to solve such a problem, various techniques for estimating a host vehicle position on a map with higher accuracy have been studied. For example, PTL 1 discloses a technique of recognizing a landmark (a landmark which is a sign having a predetermined feature) present on a map from images photographed using a camera and estimating a host vehicle position on the map.
A self-position estimating device disclosed in PTL 1 includes landmark detecting means that detects a landmark, reliability acquiring means that acquires landmark reliability which is reliability of the landmark detected by the landmark detecting means, and self-position estimating means that estimates a self-position on the basis of a plurality of landmarks detected by the landmark detecting means, and the self-position estimating means sequentially selects the landmarks to be used to estimate the self-position from the landmark having highest landmark reliability.
According to the self-position estimating device disclosed in PTL 1, when a plurality of landmarks are extracted, it is possible to accurately recognize different landmarks and to estimate a self-position with high accuracy.